KovAPIKovAPI
Image ● Operational Service status →

Seedream 5.5

Beta ⧉
seedream-5.5

Seedream 5.5 is ByteDance's newest Seedream image model, released in November 2026 (Beta). One model handles text-to-image and editing: an edit keeps the original composition and changes only what the prompt asks for, and the parts you did not ask to change stay put even after many rounds of edits. Up to 10 reference images can be fused into one picture (people, products, scenes). In-image text, Chinese included, comes out legible. 1K / 2K and 9 aspect ratios (21:9 and 9:21 included); you get the original PNG with its C2PA content credential, never cropped or resized. OpenAI-compatible images API, billed per image, failures refunded in full.

Text-to-imageImage editingMulti-image fusion1K/2KOriginal output

Live Test · Playground

Try out Seedream 5.5 right here (available after login).

Model Type

Input

Image

Output

image
Results will appear here after running.
🕑 Generated images and videos are kept for 7 days and then deleted automatically. Download anything you want to keep.

About Seedream 5.5

Seedream 5.5 is ByteDance's newest Seedream image model, released in November 2026 (Beta), one model for text-to-image and editing. An edit keeps the original composition and changes only what you ask for, even after many rounds; up to 10 reference images can be fused into one picture. In-image text, Chinese included, comes out legible. 1K / 2K in 9 ratios; you get the original PNG with its C2PA content credential, never cropped or resized. Billed per image, failures refunded in full.

Use cases

Posters and text layouts

Headlines, price tags and menus come out legible, Chinese included; 21:9 suits site banners, 9:21 long mobile images.

Edits round after round

Start from one version, then change the background, colours or a sign round by round: what you did not ask to change stays put. Good for tuning a product shot.

Multi-image fusion

Put people, products and places (up to 10 references) into one picture; say in the prompt what each image is for, in order.

How to choose

Pick it for the newest model, edits that hold up over many rounds, tall 9:21 images, or a lower price per 1K / 2K image; for batch work that needs steady throughput start with Seedream 5.0 Pro; for 4K, transparent backgrounds or GIF use Nano Banana or GPT Image 2.6.

FAQ

Which ratios and resolutions?
1K and 2K, in 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 9:21. You get that ratio's native size (2K 16:9 is 2816×1584, 2K 21:9 is 3136×1344), never cropped or resized; any other ratio returns 400 at submit and nothing is charged.
How many reference images?
Up to 10: one in image, several in images, as an image URL, a data: URI or base64. With several, say in the prompt what each one is for, in order.
What format comes back?
The original PNG with its C2PA content credential. JPEG, transparent backgrounds and layer splitting are not offered yet, and each request makes one image.
Why is it marked Beta?
Seedream 5.5 was released only recently and the service is still scaling up, so at peak times a request may fail or slow down. A failure costs nothing and you can resubmit as is; for batch work that needs steady throughput, Seedream 5.0 Pro is the safer choice for now.
How long, and what if it fails?
Usually 20–40 seconds when the service is not busy; 2K and several references take a little longer, and peaks can be slower. A failed render or a content-check rejection costs nothing: the hold is released in full.

Related Models

Explore other models you can integrate.

View all →